Leverage Dynamics and the Burden of Debt
| Author | Mathias Drehmann,Mikael Juselius |
| Published date | 01 April 2020 |
| DOI | http://doi.org/10.1111/obes.12330 |
| Date | 01 April 2020 |
347
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 82, 2 (2020) 0305–9049
doi: 10.1111/obes.12330
Leverage Dynamics and the Burden of Debt*
Mikael Juselius† and Mathias Drehmann‡
†Bank of Finland, Helsinki, Finland (e-mail: mikael.juselius@bof.fi)
‡Bank for International Settlements, Basel, Switzerland (e-mail: mathias.drehmann@bis.org)
Abstract
In addition to leverage, the debt service burden of households and firms is an important link
between financial and real developments at the aggregate level. Using US data from 1985
to 2017, we find that the debt service burden has sizeable negative effects on expenditure.
Its interplay with leverage also explains several data puzzles, including the lack of above-
trend output growth during credit booms and the severity of ensuing recessions, without
appealing to large shocks or nonlinearities. Estimating the model with data up to 2005, it
predicts credit and expenditure paths that closely match actual developments before and
during the Great Recession.
I. Introduction
Credit and output often diverged substantially in the United States over the last decades in
contrast to the intuition from standard macro-finance models. For instance, the recoveries
after the Great Recession and the recession in the early 1990s were largely ‘creditless’, i.e.
output picked up without an equal pick up in credit. Less noted but more puzzling, the
preceding booms wereessentially ‘growthless’ as credit grew rapidlywhile output remained
roughly on trend. Hence, the credit-to-GDP ratio surged ahead of the tworecessions and fell
during the recovery phases.We showthat understanding the drivers of these puzzling credit-
to-GDP developments helps to clarify the link between financial and real developments.
The importance of financial developments for future real outcomes has been docu-
mented in several empirical studies. For one, recessions that follow periods of strong
credit-to-GDP growth tend to be much deeper than otherwise (e.g. Claessens, Kose and
Terrones, 2012 or Jorda, Schularick and Taylor, 2013). More generally, medium run in-
creases in credit to GDP have a positive impact on growth in the short, but a negative one
in the medium run (Mian, Sufi and Verner 2017). But the propagation mechanism behind
*Wethank Enisse Khar roubi, Christian Upper and KostasTsatsaronis for long and helpful discussions as well as
Anton Korinek and Martin Summer for extensivecomments on earlier drafts. We are also grateful for comments from
Klaus Adam, Pierre-Richard Agenor, Claudio Borio, Fabio Canova, Giuseppe Cavaliere, Luca De Angelis, Carlo
Favero,Michael Funke, Charles Goodhart, Julia Giese, Enrique Mendoza, Chris Sims, Hyun Shin, Nikola Tarashev,
Anders Verdin and seminar participants at the Austrian Central Bank, the Board of Governors at the FederalReserve
System, Dankmarks Nationalbank, the 30th Annual Congress of the European Economic Association, the Norges
Bank, the Swiss National Bank and the first BIS Research Network Meeting. Views expressed reflect those of the
authors and not those of the Bank of Finland or the Bank for International Settlements.
348 Bulletin
these delayed real effects, and why credit and output at times diverge, is not yet fully un-
derstood. These features are hard to explain with financial accelerator-type models that
emphasize net worth or leverage (e.g. Kiyotaki and Moore, 1997; Bernanke, Gertler and
Gilchrist 1999), as they typically imply a high degree of synchronicity between real and
financial developments. More recent theoretical work highlights the role of debt service:
Following a credit boom it can constrain aggregate demand under some conditions such
as the zero lower bound (e.g. Eggertsson and Krugman, 2012; Farhi and Werning 2016
or Korinek and Simsek 2016). This suggests that the link from leverage to credit booms
and subsequent debt service burdens may be important for understanding the puzzling
behaviour of credit and output in the data.
To shed light on financial-real interactions, we empirically model the leverage and the
debt service burden of households and firms at the aggregate level to assess how they affect
real outcomes. We start from two weak empirical hypotheses. First, if leverage is constant
in the long run, it implies that the credit-to-GDP ratio is cointegrated with real asset prices.
Second, if the debt service burden is constant in the long run, it implies a cointegration
relationship between credit-to-GDP and lending rates.
We then test for and find that these two long-run relationships are in the data. Specifi-
cally,estimating a vector auto-regressive (VAR) model in error correction form on quarterly
US data from 1985 to 2017, we find that the credit-to-GDP ratio is cointegrated with real
asset prices, on the one hand, and with lending rates, on the other. These results imply
that the trend increase in the credit-to-GDP ratio over the last 30 years can be attributed to
falling lending rates and rising real asset prices. As such, the latter two variables are also
inversely related in the long-run.
More importantly, we find that the deviations from the two long-run relationships –
the leverage gap and the debt service gap henceforth – have sizeable effects on credit
and output. In line with micro-evidence and standard macro-finance models, we show that
real credit growth increases when the leverage gap is negative, for instance due to high
asset prices. And higher credit growth in turn boosts output growth. Going beyond the
existing evidence, we find that the debt service gap plays an additional important role at
the aggregate level that has generally been overlooked: it has a strong negative impact on
consumption and investment.1In addition, it negatively affects credit and real asset price
growth.
These dynamics provide a more structural interpretation of the link between credit-
to-GDP ratios and real outcomes. For instance, they explain that rapidly rising credit-to-
GDP ratios during credit booms (e.g. Mendoza and Terrones 2008 or Reinhart and Rogoff
2009) and creditless recoveries afterwards (e.g. Abiad, Li and Dell’Ariccia 2011) arise
from diverging pressures from the leverage and debt service gaps. Equally, the deep and
protracted recessions, which tend to followcredit booms, result not only from high leverage
depressing collateral values and thus credit growth, but also from households and firms
cutting back expenditure to reduce high debt service burdens.
The adjustment dynamics to the leverage and the debt service gaps are so pronounced
that they help anticipate the Great Recession. Starting from pseudo real-time estimates of
1The negative effect of debt service on expenditure is confirmed for a panel of 16 countries in follow up workby
Drehmann, Juselius and Korinek (2018).
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial